Spectrum Sensing Method Based on Wavelet Transform and Residual Network
Zhen Pan, Bangning Zhang, Zhibo Chen, Daoxing Guo, Wenfeng Ma
Abstract
Spectrum sensing is the premise and foundation of cognitive radio technology to solve the scarcity of spectrum resources. In this letter, a spectrum sensing method (WT-ResNet) based on wavelet transform (WT) and residual network (ResNet) is proposed for the problem of spectrum sensing of non-stationary signals with a low signal-to-noise ratio (SNR). The method uses the wavelet transform to process the signal for the characteristics of non-stationary signals, obtains the time-frequency matrix of the signal as the test statistic, and combines the residual network. The method is data-driven and does not require any a priori information about the signal. In addition, we not only test the commonly used wavelet functions but also analyze the SNR robustness and generalization ability of the algorithm. Finally, the simulation results show that the method outperforms other spectrum sensing methods. In particular, the method achieves a detection probability of 85% at SNR = −16dB when the false alarm probability is 10%.